检索结果-内蒙古大学图书馆

VIPLFaceNet： an open source deep face recognition SDK

Frontiers of Computer Science 2017年第2期11卷 208-218页

作者： Xin LIU Meina KAN Wanglong WU Shiguang SHAN Xilin CHEN Key Lab of Intelligent Information Processing of Chinese Academy of Sciences (CAS) Institute of Computing Technology CAS Beijing 100190 China University of Chinese Academy of Sciences Beijing 100049 China

Robust face representation is imperative to highly accurate face recognition. In this work, we propose an open source face recognition method with deep representation named as VIPLFaceNet, which is a lO-layer deep convolu- tional neural network with seven convolutional layers and three fully-connected layers. Compared with the well-known AlexNet, our VIPLFaceNet takes only 20% training time and 60% testing time, but achieves 40% drop in error rate on the real-world face recognition benchmark LFW. Our VIPLFaceNet achieves 98.60% mean accuracy on LFW us- ing one single network. An open-source C＋＋ SDK based on VIPLFaceNet is released under BSD license. The SDK takes about 150ms to process one face image in a single thread on an i7 desktop CPU. VIPLFaceNet provides a state-of-the-art start point for both academic and industrial face recognition applications.

关键词： deep learning face recognition open source,VIPLFaceNet

来源：评论

学校读者我要写书评

暂无评论

Relation Reconstructive Binarization of word embeddings

引用

Frontiers of Computer Science 2022年第2期16卷 47-54页

作者： Feiyang PAN Shuokai LI Xiang AO Qing HE Key Lab of Intelligent Information Processing of Chinese Academy of Sciences(CAS) Institute of Computing TechnologyCASBeijing 100190China University of Chinese Academy of Sciences Beijing 100049China

Word-embedding acts as one of the backbones of modern natural language processing(NLP).Recently,with the need for deploying NLP models to low-resource devices,there has been a surge of interest to compress word embeddings into hash codes or binary vectors so as to save the storage and memory ***,existing work learns to encode an embedding into a compressed representation from which the original embedding can be *** these methods aim to preserve most information of every individual word,they often fail to retain the relation between words,thus can yield large loss on certain *** this end,this paper presents Relation Reconstructive Binarization(R2B)to transform word embeddings into binary codes that can preserve the relation between *** its heart,R2B trains an auto-encoder to generate binary codes that allow reconstructing the wordby-word relations in the original embedding *** showed that our method achieved significant improvements over previous methods on a number of tasks along with a space-saving of up to 98.4%.Specifically,our method reached even better results on word similarity evaluation than the uncompressed pre-trained embeddings,and was significantly better than previous compression methods that do not consider word relations.

关键词： embedding compression variational auto-encoder binary word embedding

来源：评论

学校读者我要写书评

暂无评论

Nested relation extraction with iterative neural network

引用

Frontiers of Computer Science 2021年第3期15卷 109-122页

作者： Yixuan CAO Dian CHEN Zhengqi XU Hongwei LI Ping LUO Key Lab of Intelligent Information Processing of Chinese Academy of Sciences(CAS) Institute of Computing TechnologyCASBeijing 100190China University of Chinese Academy of Sciences Beijing 100049China

Most existing researches on relation extraction focus on binary flat relations like Bomln relation between a Person and a *** a large portion of objective facts de-scribed in natural language are complex,especially in professional documents in fields such as finance and biomedicine that require precise *** example,“the GDP of the United States in 2018 grew 2.9%compared with 2017”describes a growth rate relation between two other relations about the economic index,which is beyond the expressive power of binary flat ***,we propose the nested relation extraction problem and formulate it as a directed acyclic graph(DAG)structure extraction ***,we propose a solution using the Iterative Neural Network which extracts relations layer by *** proposed solution achieves 78.98 and 97.89 FI scores on two nested relation extraction tasks,namely semantic cause-and-efFect relation extraction and formula ***,we observe that nested relations are usually expressed in long sentences where entities are mentioned repetitively,which makes the annotation difficult and ***,we extend our model to incorporate a mention-insensitive mode that only requires annotations of relations on entity concepts(instead of exact mentions)while preserving most of its *** mention-insensitive model performs better than the mention sensitive model when the random level in mention selection is higher than 0.3.

关键词： nested relation extraction mention insensitive relation iterative neural network

来源：评论

学校读者我要写书评

暂无评论

Robust low-rank tensor factorization by cyclic weighted median

引用

Science China(information Sciences) 2015年第5期58卷 145-155页

作者： MENG DeYu ZHANG Biao XU ZongBen ZHANG Lei GAO ChenQiang Institute for Information and System Sciences and Ministry of Education Key Lab for Intelligent Networks and Network Security Xi'an Jiaotong University Department of Computing Hong Kong Polytechnic University Chongqing Key Laboratory of Signal and Information Processing Chongqing University of Posts and Telecommunications

Low-rank tensor factorization(LRTF) provides a useful mathematical tool to reveal and analyze multi-factor structures underlying data in a wide range of practical applications. One challenging issue in LRTF is how to recover a low-rank higher-order representation of the given high dimensional data in the presence of outliers and missing entries, i.e., the so-called robust LRTF problem. The L1-norm LRTF is a popular strategy for robust LRTF due to its intrinsic robustness to heavy-tailed noises and outliers. However, few L1-norm LRTF algorithms have been developed due to its non-convexity and non-smoothness, as well as the high order structure of data. In this paper we propose a novel cyclic weighted median(CWM) method to solve the L1-norm LRTF problem. The main idea is to recursively optimize each coordinate involved in the L1-norm LRTF problem with all the others fixed. Each of these single-scalar-parameter sub-problems is convex and can be easily solved by weighted median filter, and thus an effective algorithm can be readily constructed to tackle the original complex problem. Our extensive experiments on synthetic data and real face data demonstrate that the proposed method performs more robust than previous methods in the presence of outliers and/or missing entries.

关键词： tensor factorization low-rank weighted median filter tensorface

来源：评论

学校读者我要写书评

暂无评论

A Regularized Super Resolution Algorithm for Generalized Gaussian Noise

引用

Journal of Donghua University(English Edition) 2010年第1期27卷 25-35页

作者：陈文方向忠刘立峰蒋伟丁大为乔艳涛 Institute of Image Communication and Information Processing Shanghai Jiaotong University Computer and Information Department Shanghai University of Electric Power Key Lab oratory of Intelligent Computing & Signal Processing Ministry of EducationAnhui University

In this paper,an iterative regularized super resolution (SR) algorithm considering non-Gaussian noise is *** on the assumption of a generalized Gaussian distribution for the contaminating noise,an lp norm is adopted to measure the data fidelity term in the cost *** the meantime,a regularization functional defined in terms of the desired high resolution (HR) image is employed,which allows for the simultaneous determination of its value and the partly reconstructed image at each iteration *** convergence is thoroughly *** results show the effectiveness of the proposed algorithm as well as its superiority to conventional SR methods.

关键词： super resolution generalized p-Gaussian distribution regularization parameter

来源：评论

学校读者我要写书评

暂无评论

Localization Optimization Algorithm of Maximum Likelihood Estimation Based on Received Signal Strength 9

Localization Optimization Algorithm of Maximum Likelihood Es...

引用

2017 9th IEEE International Conference on Communication Software and Networks (ICCSN 2017)

作者： Jun Zhu Qiang Chen Jingjing Zhang Key Lab of Intelligent Computing and Signal Processing Ministry of Education School of Electronics and Information Engineering Anhui University

ISBN: (纸本)9781509038237;9781509038220

In this paper, we propose a method to improve localization algorithm of maximum likelihood estimation;the localization scheme relies on the distance threshold. In order to suppress effectively the effects of received signal strength error to node localization precision. This paper presents an indoor localization algorithm based on received signal strength to select anchor nodes. Compared with the traditional localization algorithm, this scenario not only improve the localization accuracy, but also reduce the calculation complexity of nodes. The simulation results show that the average error of the proposed method is less than 0.15 m. Moreover, when there are a large number of anchor nodes, the computational complexity is effectively reduced. Verification result verifies the effectiveness and reliability of the algorithm.

关键词： localization received signal strength(RSS) maximum likelihood estimation(MLE) least squares threshold

来源：评论

学校读者我要写书评

暂无评论

Speaker Recognition Based on the Improved Double-Threshold Endpoint Algorithm and Multistage Vector Quantization 9

Speaker Recognition Based on the Improved Double-Threshold E...

引用

2017 9th IEEE International Conference on Communication Software and Networks (ICCSN 2017)

作者： Jun Zhu Jingjing Zhang Qiang Chen Peipei Tu Key Lab of Intelligent Computing and Signal Processing Ministry of Education School of Electronics and Information Engineering Anhui University

ISBN: (纸本)9781509038237;9781509038220

The traditional double-threshold endpoint detection method has the phenomenon of missing detection. Therefore, the speech recognition(SR) system based on vector quantization(VQ) in this paper proposes an improved algorithm for this phenomenon, which effectively avoids the problem of missing detection. Then, Mel Frequency Cepstral Coefficients(MFCC) is used to extract the characteristic parameters of the speech signal, and the multistage vector quantization is used to quantify the characteristic parameters. Experimental results show that, the proposed algorithm improves the recognition rate of the text-independent speaker recognition system by 8.7%, and it also confirms that the longer the training speech is, the higher the recognition rate will be.

关键词： the multistage vector quantization MFCC double-threshold endpoint detection method missing detection characteristic parameters

来源：评论

学校读者我要写书评

暂无评论

Semantic-driven management and search for resources in research community

Semantic-driven management and search for resources in resea...

引用

2006 2nd International Conference on Semantics Knowledge and Grid, SKG

作者： Junsheng, Zhang Zhixian, Yan Key Lab. of Intelligent Information Processing Institute of Computing Technology Chinese Academy of Sciences

ISBN: (纸本)0769532055

In recent years, semantic search has become one hot motivation of the semantic web. In this paper, we propose a semantic-based resource management and search architecture and its implementation in research community, including most of the research relevant resources. We have built two authoring tools respectively for concepts and instances. With the authoring tools, semantic search portal is provided, which has three different interfaces, respectively for retrieval of concept, instance and semantic relationship. Compared with the traditional search methods, semantic search methodology not only gives us more interesting and satisfying search results, whose precision is higher than those of traditional ones only based on keywords, but also displays the results in a more structural and semantic way. © 2006 IEEE.

关键词：

来源：评论

学校读者我要写书评

暂无评论

A resource space model for dataspace

A resource space model for dataspace

引用

6th International Conference on Semantics, Knowledge and Grid, SKG 2010

作者： Jiang, Xiaorui Sun, Xiaoping Zhuge, Hai Knowledge Grid Research Group Key Lab of Intelligent Information Processing Institute of Computing Technology CAS Beijing China

ISBN: (纸本)9780769541891

Data space is a semi-structural data model for the management of large-scale heterogeneous data objects. In a data space, each data object consists of a set of attribute-value pairs to describe the internal properties of the object and their relationships to other objects. The data space model provides a flexible query language that supports attribute queries and semantic link queries. In this paper we introduce a Resource Space Model that extends the concepts of data space to make it more flexible by incorporating classification semantics. We model a data space as a set of Resource Spaces (RS) with each RS representing a kind of resources that share common attributes and can be put into the same category, either according to a user's knowledge of the world or his query preferences. Our model is suitable for designing a user-tailored semantic view of a data space while at the same time without losing the schema-later data-centric nature. A practical query language is introduced to implement flexible query operations on the resource spaces of the data space. Examples show how the proposed model and language support both the schema query and the data query. © 2010 IEEE.

关键词： Query languages

来源：评论

学校读者我要写书评

暂无评论

Merging complex networks

Merging complex networks

引用

7th International Conference on Semantics, Knowledge, and Grids, SKG 2011

作者： Sun, Xiaoping Zhuge, Hai Key Lab of Intelligent Information Processing Institute of Computing Technology Chinese Academy of Sciences China

ISBN: (纸本)9780769545158

In the cyber-physical society, networks are constructed for information transportation. Among them, power law networks with the scale free property are extensively found in self-organized systems. The dynamicity of the cyber-physical society drives large-scale networks keeping interacting and evolving. Several networks can be integrated or merged into one network during the evolving process of the whole cyber-physical society either because they share the same nodes or because one network is trying to connect to the other. A natural question is that in what way two or more networks will merge with each other so that their previous scale-free properties still hold. In this paper, we conducted a set of simulation experiments to study the effects of different merging processes on the degree distribution of merged networks. The result can be used to understand the merging process of complex networks in the cyber-physical society and also can be used to design an integration strategy for multiple networks. © 2011 IEEE.

关键词： Complex networks

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：